Select short name of variable defining ModelVerticalGrid

[pchakraborty]

Types of change(s)

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Trivial change (affects only documentation or cleanup)
• Refactor (no functional changes, no api changes)

Checklist

• Tested this change with a run of GEOSgcm
• Ran the Unit Tests (make tests)

Description

• A Model VerticalGrid is now instantiated with standard_name, units and num_levels. The method add_short_name is used to specify variable short names corresponding to edge and center vertical staggerings, which is selected at runtime using the method get_short_name.
• Updated ComponentSpecParse::parse_geometry_spec to use the new interface.
• Updated Test_ModelVerticalGrid.pf to use the new interface.
• Updated vertical regridding tests to specify standard_name instead of short_name for Model VerticalGrids.
• Cleanup of FixedLevels VerticalGrid - we didn't need MAPL_BaseMod::MAPL_GridGet after all.

Related Issue

[tclune]

You need an "unusable" argument here to separate out the options. Both arguments are the same type (character), so it is dangerously ambiguous which one is intended.

[pchakraborty]

Not sure I understand. Both edge and center are optional arguments. Did you mean something like

   subroutine add_short_names(this, edge, unusable, center)
      class(ModelVerticalGrid), intent(inout) :: this
      character(*), optional, intent(in) :: edge
      class(KeywordEnforder), optional, intent(in) :: unusable
      character(*), optional, intent(in) :: center

      if (present(edge)) this%short_name_edge = edge
      if (present(center)) this%short_name_center = center
      _UNUSED_DUMMY(unusable)
   end subroutine add_short_names

[tclune]

No - I mean "unusable" should be even before edge. Force the user to be explicit since there is no obvious ordering of "center" vs "edge" in this context.

[tclune]

To be explicit, consider this call:

call obj%add_short_name('x')

Is it obvious that this was adding "edge"?
Or consider:

call obj%add_short_name('x', 'y')

Which is which? Instead:

call obj%add_short_name(edge='x', center='y')

is unambiguous.

[pchakraborty]

Got it. I had seen the unusable argument, but was not familiar with its usage. Pushed the change.

[tclune]

I picked up the technique from ESMF. They use it to enforce the rule that interfaces remain backward compatible. All new arguments must be optional and after the "unusable" argument. But the same technique also allows us to enforce using keywords for ambiguous situations. At one point I was going to rigorously enforce that all MAPL optional arguments be after "unusable" as a stylistic statement that makes optional arguments more obviously optional. But am no longer so dogmatic. Becomes a pain to put "unusable" in interfaces that just have optional "rc".

[tclune]

A couple of minor changes requested. See inline comments.

[tclune]

Looks good.